>>
>> Sometimes I suffered the nat_tree_lock contention between
>> f2fs_write_checkpoint
>> and f2fs_get_node_info. Commit a9419b6("f2fs: do not bother checkpoint by
>> f2fs_get_node_info") also mentioned that situation.
>>
>> My idea is, when flush nat entries, we can use some structures to record nat
>> pages we may read, and readahead them before hold nat_tree_lock. Before
>> impletement code, I did some survey and found a submittion in community.
>>
>> Subject: f2fs: use bucket sort to avoid tree lookup and list sort when nat
>> flushing
>> Link:
>> https://lore.kernel.org/linux-f2fs-devel/20170520122435.17574-2-houpengy...@huawei.com/
>> This patch aims to improve nat entry set sort by using bucket.
>> I steal that structure and readahead nat pages contain nat entry set which
>> cannot be moved
>> to journal according to dirty nat entry set bucket.
>>
>> By doing this, I think there are two benefits to reducing nat_tree_lock hold
>> time when
>> when flush nat entries.
>> 1. avoid nat set tree lookup and sort
>> 2. readahead nat pages before holding nat_tree_lock
>>
>> Signed-off-by: wangzijie <wangzij...@honor.com>
>> ---
>> fs/f2fs/f2fs.h | 1 +
>> fs/f2fs/node.c | 70 ++++++++++++++++++++++++--------------------------
>> fs/f2fs/node.h | 2 +-
>> 3 files changed, 35 insertions(+), 38 deletions(-)
>>
>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>> index 46be75605..b27cc059f 100644
>> --- a/fs/f2fs/f2fs.h
>> +++ b/fs/f2fs/f2fs.h
>> @@ -975,6 +975,7 @@ struct f2fs_nm_info {
>> struct radix_tree_root nat_set_root;/* root of the nat set cache */
>> struct f2fs_rwsem nat_tree_lock; /* protect nat entry tree */
>> struct list_head nat_entries; /* cached nat entry list (clean) */
>> + struct list_head nat_dirty_set[NAT_ENTRY_PER_BLOCK + 1]; /*
>> store dirty nat set */
>> spinlock_t nat_list_lock; /* protect clean nat entry list */
>> unsigned int nat_cnt[MAX_NAT_STATE]; /* the # of cached nat entries
>> */
>> unsigned int nat_blocks; /* # of nat blocks */
>> diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
>> index 27743b93e..87c975ee8 100644
>> --- a/fs/f2fs/node.c
>> +++ b/fs/f2fs/node.c
>> @@ -244,6 +244,12 @@ static void __del_from_nat_cache(struct f2fs_nm_info
>> *nm_i, struct nat_entry *e)
>> __free_nat_entry(e);
>> }
>>
>> +static void __relocate_nat_entry_set(struct f2fs_nm_info *nm_i,
>> + struct nat_entry_set
>> *set)
>> +{
>> + list_move_tail(&set->set_list, &nm_i->nat_dirty_set[set->entry_cnt]);
>> +}
>Hi zijie,
>Is there any lock protecting this list related operations in the
>current process?
Hi, Zhiguo
I think the lock protection needed is as same as __set_nat_cache_dirty().
>> +
>> static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i,
>> struct nat_entry *ne)
>> {
>> @@ -260,6 +266,7 @@ static struct nat_entry_set *__grab_nat_entry_set(struct
>> f2fs_nm_info *nm_i,
>> head->set = set;
>> head->entry_cnt = 0;
>> f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head);
>> + __relocate_nat_entry_set(nm_i, head);
>> }
>> return head;
>> }
>> @@ -279,8 +286,10 @@ static void __set_nat_cache_dirty(struct f2fs_nm_info
>> *nm_i,
>> * 2. update old block address to new one;
>> */
>> if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) ||
>> - !get_nat_flag(ne, IS_DIRTY)))
>> + !get_nat_flag(ne, IS_DIRTY))) {
>> head->entry_cnt++;
>> + __relocate_nat_entry_set(nm_i, head);
>> + }
>>
>> set_nat_flag(ne, IS_PREALLOC, new_ne);
>>
>> @@ -309,6 +318,7 @@ static void __clear_nat_cache_dirty(struct f2fs_nm_info
>> *nm_i,
>>
>> set_nat_flag(ne, IS_DIRTY, false);
>> set->entry_cnt--;
>> + __relocate_nat_entry_set(nm_i, set);
>> nm_i->nat_cnt[DIRTY_NAT]--;
>> nm_i->nat_cnt[RECLAIMABLE_NAT]++;
>> }
>> @@ -2976,24 +2986,6 @@ static void remove_nats_in_journal(struct
>> f2fs_sb_info *sbi)
>> up_write(&curseg->journal_rwsem);
>> }
>>
>> -static void __adjust_nat_entry_set(struct nat_entry_set *nes,
>> - struct list_head *head, int
>> max)
>> -{
>> - struct nat_entry_set *cur;
>> -
>> - if (nes->entry_cnt >= max)
>> - goto add_out;
>> -
>> - list_for_each_entry(cur, head, set_list) {
>> - if (cur->entry_cnt >= nes->entry_cnt) {
>> - list_add(&nes->set_list, cur->set_list.prev);
>> - return;
>> - }
>> - }
>> -add_out:
>> - list_add_tail(&nes->set_list, head);
>> -}
>> -
>> static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
>> const struct f2fs_nat_block *nat_blk)
>> {
>> @@ -3095,6 +3087,7 @@ static int __flush_nat_entry_set(struct f2fs_sb_info
>> *sbi,
>>
>> /* Allow dirty nats by node block allocation in write_begin */
>> if (!set->entry_cnt) {
>> + list_del(&set->set_list);
>> radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
>> kmem_cache_free(nat_entry_set_slab, set);
>> }
>> @@ -3109,11 +3102,8 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi,
>> struct cp_control *cpc)
>> struct f2fs_nm_info *nm_i = NM_I(sbi);
>> struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
>> struct f2fs_journal *journal = curseg->journal;
>> - struct nat_entry_set *setvec[NAT_VEC_SIZE];
>> struct nat_entry_set *set, *tmp;
>> - unsigned int found;
>> - nid_t set_idx = 0;
>> - LIST_HEAD(sets);
>> + int i;
>> int err = 0;
>>
>> /*
>> @@ -3129,6 +3119,16 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi,
>> struct cp_control *cpc)
>> if (!nm_i->nat_cnt[DIRTY_NAT])
>> return 0;
>>
>> + /* readahead sets which cannot be moved to journal */
>> + if (!__has_cursum_space(journal, nm_i->nat_cnt[DIRTY_NAT],
>> NAT_JOURNAL)) {
>> + for (i = MAX_NAT_JENTRIES(journal); i <=
>> NAT_ENTRY_PER_BLOCK; i++) {
>i am a little confused, why is there "i <= NAT_ENTRY_PER_BLOCK;"?
>do we need to calculate according to the current nat block/entry number?
>thanks!
Yes, you are right, we can calculate according to DIRTY_NAT count to end this
loop early.
>> + list_for_each_entry_safe(set, tmp,
>> &nm_i->nat_dirty_set[i], set_list) {
>> + f2fs_ra_meta_pages(sbi, set->set, 1,
>> + META_NAT,
>> true);
>> + }
>> + }
>> + }
>> +
>> f2fs_down_write(&nm_i->nat_tree_lock);
>>
>> /*
>> @@ -3141,21 +3141,13 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi,
>> struct cp_control *cpc)
>> nm_i->nat_cnt[DIRTY_NAT], NAT_JOURNAL))
>> remove_nats_in_journal(sbi);
>>
>> - while ((found = __gang_lookup_nat_set(nm_i,
>> - set_idx, NAT_VEC_SIZE, setvec))) {
>> - unsigned idx;
>> -
>> - set_idx = setvec[found - 1]->set + 1;
>> - for (idx = 0; idx < found; idx++)
>> - __adjust_nat_entry_set(setvec[idx], &sets,
>> - MAX_NAT_JENTRIES(journal));
>> - }
>> -
>> /* flush dirty nats in nat entry set */
>> - list_for_each_entry_safe(set, tmp, &sets, set_list) {
>> - err = __flush_nat_entry_set(sbi, set, cpc);
>> - if (err)
>> - break;
>> + for (i = 0; i <= NAT_ENTRY_PER_BLOCK; i++) {
>> + list_for_each_entry_safe(set, tmp, &nm_i->nat_dirty_set[i],
>> set_list) {
>> + err = __flush_nat_entry_set(sbi, set, cpc);
>> + if (err)
>> + break;
>> + }
>> }
>>
>> f2fs_up_write(&nm_i->nat_tree_lock);
>> @@ -3249,6 +3241,7 @@ static int init_node_manager(struct f2fs_sb_info *sbi)
>> struct f2fs_nm_info *nm_i = NM_I(sbi);
>> unsigned char *version_bitmap;
>> unsigned int nat_segs;
>> + int i;
>> int err;
>>
>> nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
>> @@ -3275,6 +3268,9 @@ static int init_node_manager(struct f2fs_sb_info *sbi)
>> INIT_LIST_HEAD(&nm_i->nat_entries);
>> spin_lock_init(&nm_i->nat_list_lock);
>>
>> + for (i = 0; i <= NAT_ENTRY_PER_BLOCK; i++)
>> + INIT_LIST_HEAD(&nm_i->nat_dirty_set[i]);
>> +
>> mutex_init(&nm_i->build_lock);
>> spin_lock_init(&nm_i->nid_list_lock);
>> init_f2fs_rwsem(&nm_i->nat_tree_lock);
>> diff --git a/fs/f2fs/node.h b/fs/f2fs/node.h
>> index 030390543..d805d4ce7 100644
>> --- a/fs/f2fs/node.h
>> +++ b/fs/f2fs/node.h
>> @@ -158,7 +158,7 @@ enum mem_type {
>> };
>>
>> struct nat_entry_set {
>> - struct list_head set_list; /* link with other nat sets */
>> + struct list_head set_list; /* link with nat sets which have
>> same entry_cnt */
>> struct list_head entry_list; /* link with dirty nat entries */
>> nid_t set; /* set number*/
>> unsigned int entry_cnt; /* the # of nat entries in set */
>> --
>> 2.25.1
By the way, Chao and Zhiguo
I noticed that in f2fs_flush_nat_entries(), the check for nat_cnt[DIRTY_NAT] is
out of
nat_tree_lock, and in f2fs_write_checkpoint(), there has below check without
nat_tree_lock:
if (cpc->reason & CP_DISCARD) {
if (!f2fs_exist_trim_candidates(sbi, cpc)) {
unblock_operations(sbi);
goto out;
}
if (NM_I(sbi)->nat_cnt[DIRTY_NAT] == 0 &&
SIT_I(sbi)->dirty_sentries == 0 &&
prefree_segments(sbi) == 0) {
f2fs_flush_sit_entries(sbi, cpc);
f2fs_clear_prefree_segments(sbi, cpc);
unblock_operations(sbi);
goto out;
}
}
Is there a missing or we don't need nat_tree_lock in these situations?
And do you think the way in this patch is appropriate for reducing nat_tree_lock
hold time?
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
